Predicting COVID-19 Case Count Based on Prior Factors

DS 3000 Summer 1 2020 Final Project

By Luca Demian and Max Breslauer-Friedman

This is our repository for all our final project stuff.

Useful Links

• Data Acquisition: /data/
• Final Report: /processing/DS3000_FP4_Group25.ipynb

Description

For our project, we wanted to do something using COVD-19 data. At first, we were thinking of different unsupervised project ideas about analyzing the factors which contributed to the pandemic spread in the United States. After review we switched to thinking of supervised learning problems, and decided to attempt to develop a model to predict COVID-19 case count in a particular county based on several prior factors we identified. We decided to look at case count N days after the first case recorded, to control for how the infection appeared non-uniformly throughout the states. We are going to examine several different timelines, or N-day values.

Data Sources

Data Source	Link(s)	Date	Features
NY Times	Link	May 22, 2020	date_0days, cases_0days, deaths_0days, date_90days, cases_90days, deaths_90days, fips
Kaiser Health	Link	March 30, 2020	icu_beds, kaiser_total_population, kaiser_60plus_population
US Census Commuting Data	Link	2015	commuting_within, commuting_out, commuting_in
US Census Population Density	Link	2019	density, pop_2019
Political Majority	Link	2016	percent_democrat, percent_gop
Poverty Data	Link	Feb. 5, 2020	percent_in_poverty
Age Data	Link	2018	pop_under18, pop_over65, pop_2018, median_age
Education Data	Link	2018	percent_with_bachelors
Income Data	Link	2018	median_income

Selected Features

These are the features we selected to train our models on, out of the selection of features we gathered data for.

Name	Type	Description
cases_60days	Target	This is the count of cases in each county 60 days after the first recorded case.
cases_70days	Target	This is the count of cases in each county 70 days after the first recorded case.
icu_beds	Feature	Count of ICU beds per county.
commuting_in	Feature	Population commuting into this county every day.
commuting_out	Feature	Population commuting out of this county every day.
commuting_within	Feature	Population commuting within this county every day.
density	Feature	A measure of poulation density in this county.
percent_democrat	Feature	Percentage of 2016 voters voting Democratic.
percent_in_poverty	Feature	Percentage of county population (all ages) in poverty.
pop_under18	Feature	Population in county aged under 18.
pop_over65	Feature	Population in county aged 65+.
median_age	Feature	Median age for the county.
percent_with_bachelors	Feature	Percent of people in county attaining a bachelors degree or higher.
median_income	Feature	A measure of the median income in this county.
pop_2019	Feature	County population as of 2019.

Important Notes

Because the NY Times data is our base dataset, and they report all counties in New York City as a single geography, we are missing the New York City counties in our final dataset. This effects: New York, Kings, Queens, Bronx and Richmond Counties